Forking independence from the categorical point of view (Q1731575)

The model-theoretic notion of forking introduced by Shelah for classes of models axiomatized by stable first-order theories is a generalization of linear independence in vector spaces and algebraic independence in fields. In fact, (non)forking can be seen as a commutative diagram of embeddings also known as an amalgam. In this paper, the broad model-theoretic framework of abstract elementary classes (AECs) is used. A \(\mu\)-AEC is simply an accessible category with all morphisms monomorphisms. The authors describe when a category has a `stable independence notion' -- a class of distinguished commutative squares that itself forms an accessible category -- and show that this is a purely category-theoretic axiomatization of forking in a \(\mu\)-AEC. This generalizes a result of [\textit{W. Boney} et al., Ann. Pure Appl. Logic 167, No. 7, 590--613 (2016; Zbl 1400.03060)] that characterized stable forking in the framework of AECs but depended on set-representations of their objects. The category \(\mathcal K_{\mathrm{reg}}\) of regular monomorphisms in a locally presentable coregular category \(\mathcal K\) that has effective unions, in the sense of Barr, is shown to have a stable independence notion, thus showing that forking occurs in both Grothendieck toposes and Grothendieck abelian categories. Assuming a large cardinal axiom the authors also characterize when a stable independence notion exists in a \(\mu\)-AEC. Thus, it is established that model-theoretic stability is invariant under equivalence of categories.